AI-Powered Tool Assesses Understandability of Online Health Videos, Boosting Patient Engagement
A new approach leveraging artificial intelligence is poised to improve how patients access and understand crucial health information online. Despite nearly 93% of U.S. adults going online and 80% of those seeking health information, low health literacy remains a significant barrier, prompting researchers to develop a system for evaluating the clarity of patient education videos – particularly those focused on conditions like diabetes.
The study, appearing in the Journal of Medical Internet Research, details a “human-in-the-loop” augmented intelligence system designed to assess video understandability on YouTube, the world’s largest video-sharing platform. This innovative method combines established guidelines for patient education materials with machine learning and expert analysis, offering a pathway to more effective digital health communication.
“Billions of people globally are turning to social media for health information without a reliable way to confirm its accuracy, clarity, or relevance,” explains a lead researcher from Carnegie Mellon University’s Heinz College. “There’s an urgent need to curate this information, ensuring it meets the diverse health literacy needs of the population.”
The research team, comprised of experts from Carnegie Mellon University (CMU), Arizona State University (ASU), and Michigan State University (MSU), focused on diabetes – a prevalent chronic condition – collecting nearly 10,000 related YouTube videos using keywords from patient forums. The system utilizes the Patient Education Materials Assessment Tool (PEMAT) alongside features extracted directly from the videos.
The core of the system lies in a “co-training” method, where machine learning algorithms work in tandem with input from medical professionals to classify videos based on understandability. Analysis revealed a strong correlation between higher understandability scores and increased viewer engagement, measured by views, likes, and comments. Furthermore, videos deemed more understandable were more likely to receive recommendations from experts.
“Improving video understandability is critical for enhancing patient engagement with educational materials,” notes a co-author from ASU’s W.P. Carey School of Business. “This has the potential to significantly improve health literacy across individuals and communities.”
The researchers suggest the approach is adaptable to a wide range of health topics, including cardiovascular disease, cancer, medication adherence, and patient safety. However, they acknowledge limitations. The PEMAT tool, originally designed for professionally produced materials, requires modification for evaluating user-generated content like YouTube videos. Additionally, the study’s reliance on evaluations from just four physicians introduces a potential for bias.
Despite these caveats, the study highlights the potential of open platforms to provide a credible alternative to proprietary technologies in the realm of public health literacy. “Providing a strong open platform fosters innovation in data collection, digital platforms, and technologies specifically designed to address health literacy challenges,” states a researcher from Michigan State University’s Broad College of Business.
Ultimately, this research offers a promising path toward patient empowerment and improved population health by making understandable and relevant health information more readily accessible. The findings underscore the importance of prioritizing clarity and accessibility in online health content, paving the way for a more informed and engaged patient population.
More information: Xiao Liu et al, Promoting Health Literacy With Human-in-the-Loop Video Understandability Classification of YouTube Videos: Development and Evaluation Study, Journal of Medical Internet Research (2025). DOI: 10.2196/56080
